Re: [Boost-testing] git, possible redundancy in run.py

Subject: Re: [Boost-testing] git, possible redundancy in run.py
From: Niklas Angare (li51ckf02_at_[hidden])
Date: 2014-09-22 19:30:06

"Ben Pope" <benpope81_at_[hidden]> wrote:
> I'm no git or test infrastructure expert, but it looks to me like it 
> fetches submodules 3 times, once for develop and twice for master.  If I'm 
> building master, shouldn't it just fetch master submodules a single time?

Yes, it's very inefficient.

> In fact, when the pull is done, does it even matter whether master or 
> develop is checked out?

I don't know the answer to that, but using pull seems like a bad idea to me. 
What pull does is fetch and then merge remote changes with local. But there 
should never be local changes in the tester repositories. And if there are, 
they should never be merged as that would cause the tests to silently run 
with those local changes.

In fact, even using local branches (specifically, remote tracking branches) 
is unnecessary because the regression scripts will never make commits. To 
eliminate pull and local branches, you just do "git fetch" followed by "git 
checkout origin/<branch>".

> Also, if you can answer that question, perhaps you can answer this: Can I 
> run my master and develop tests from the same directory, I currently have 
> them separated, as that's how I did it with svn.

Yes, you can. But if you use the patch_boost facility, you're more likely to 
run into this bug:
https://svn.boost.org/trac/boost/ticket/10150

Regards,

Niklas Angare